聚合因子水平计数 - 按因子

Question

我一直在尝试制作一个 table 显示另一个因素的因素水平计数。为此，我查看了数十页、问题……试图使用某些包（dplyr、reshape）中的函数来完成工作，但没有成功地正确使用它们。

这就是我得到的：

# my data:
var1 <- c("red","blue","red","blue","red","red","red","red","red","red","red","red","blue","red","blue")
var2 <- c("0","1","0","0","0","0","0","0","0","0","1","0","0","0","0")
var3 <- c("2","2","1","1","1","3","1","2","1","1","3","1","1","2","1")
var4 <- c("0","1","0","0","0","0","1","0","1","1","0","1","0","1","1")
mydata <- data.frame(var1,var2,var3,var4)
head(mydata)

尝试n+1：仅显示另一个因素的因素总数。

t(aggregate(. ~ var1, mydata, sum))

      [,1]   [,2] 
var1 "blue" "red"
var2 " 5"   "12" 
var3 " 5"   "18" 
var4 " 6"   "16"

尝试 n+2：这是正确的格式，但我无法让它在多个因素上发挥作用。

library(dplyr)
data1 <- ddply(mydata, c("var1", "var3"), summarise,
            N    = length(var1))
library(reshape)
df1 <- cast(data1, var1 ~ var3, sum)
df1 <- t(df1)
df1

   blue red
1    3   6
2    1   3
3    0   2

我想要的是：

        blue red
var2.0    3  10
var2.1    1   1
var3.1    3   6
var3.2    1   3
var3.3    0   2
var4.0    2   6
var4.1    2   5

如何获得这种格式？非常感谢，

Answer 1

我们可以通过 'var1' melt 数据集，然后使用 table

library(reshape2)
tbl <- table(transform(melt(mydata, id.var="var1"),
        varN = paste(variable, value, sep="."))[c(4,1)])
names(dimnames(tbl)) <- NULL
tbl 
#
#         blue red
#  var2.0    3  10
#  var2.1    1   1
#  var3.1    3   6
#  var3.2    1   3
#  var3.3    0   2
#  var4.0    2   6
#  var4.1    2   5

或使用 dplyr/tidyr，我们将数据集从 'wide' 转换为 'long' 格式 gather，然后 unite 列（'var', 'val')创建'varV',得到'var1'和'varV'分组后的频率(tally),然后spread到'wide'格式。

library(dplyr)
library(tidyr)
gather(mydata, var, val, -var1) %>% 
           unite(varV,var, val, sep=".") %>%
           group_by(var1, varV) %>% 
           tally() %>% 
           spread(var1, n, fill = 0)
#    varV  blue   red
#   <chr> <dbl> <dbl>
#1 var2.0     3    10
#2 var2.1     1     1
#3 var3.1     3     6
#4 var3.2     1     3
#5 var3.3     0     2
#6 var4.0     2     6
#7 var4.1     2     5

聚合因子水平计数 - 按因子

Aggregating factor level counts - by factor

r

plyr

reshape

reshape2

dplyr